app.py

import gradio as gr
import pandas as pd
from openai import OpenAI
import os
import io
import json
from transformers import pipeline

# Modello NLP open-source locale per il primo livello di analisi
sentiment_pipeline = pipeline("sentiment-analysis", model="nlptown/bert-base-multilingual-uncased-sentiment")

# Funzione per analizzare i commenti con GPT solo se necessario
def analyze_comments(texts):
    """ Usa prima un modello NLP locale, poi GPT-3.5/GPT-4 solo se serve."""
    client = OpenAI(api_key=os.environ.get("OPENAI_API_KEY"))
    results = []
    to_send_gpt = []
    index_map = {}
    
    # 1️⃣ Analisi iniziale con modello open-source
    for i, text in enumerate(texts):
        local_result = sentiment_pipeline(text)[0]['label']
        rating = int(local_result.split()[0])
        
        if rating > 3:
            results.append("Positive")
        elif rating < 3:
            results.append("Negative")
        else:
            results.append("Neutral")
    
    return results


def analyze_file(file):
    """Carica un file e analizza il sentiment dei commenti nella prima colonna."""
    try:
        file_extension = os.path.splitext(file.name)[-1].lower()
        
        if file_extension == ".csv":
            df = pd.read_csv(file)
        elif file_extension in [".xls", ".xlsx"]:
            df = pd.read_excel(file, engine='openpyxl')
        else:
            return "Error: Unsupported file format.", None
        
        if df.empty or df.shape[1] < 1:
            return "Error: No valid data in the file.", None
        
        df = df.rename(columns={df.columns[0]: "Comment"})
        batch_size = 10
        sentiments = []
        
        for i in range(0, len(df), batch_size):
            batch = df['Comment'][i:i+batch_size].tolist()
            sentiments.extend(analyze_comments(batch))
        
        df["Sentiment"] = sentiments
        output = io.BytesIO()
        with pd.ExcelWriter(output, engine='openpyxl') as writer:
            df.to_excel(writer, index=False)
        output.seek(0)
        return df, output
    except Exception as e:
        return f"Error processing file: {str(e)}", None


def sentiment_analysis_interface():
    """Interfaccia Gradio per caricare file e ottenere analisi del sentiment."""
    file_input = gr.File(label="📥 Upload a file (CSV, XLSX)")
    results_output = gr.Dataframe()
    download_button = gr.File(label="📥 Download Excel")
    
    def process_file(uploaded_file):
        df, excel_data = analyze_file(uploaded_file)
        if isinstance(df, str):
            return df, None
        with open("sentiment_analysis.xlsx", "wb") as f:
            f.write(excel_data.read())
        return df, "sentiment_analysis.xlsx"
    
    return gr.Interface(
        fn=process_file,
        inputs=[file_input],
        outputs=[results_output, download_button],
        title="📊 Sentiment Analysis",
        description="Upload a file with comments and get sentiment analysis using an NLP model!"
    )

iface = sentiment_analysis_interface()

if __name__ == "__main__":
    iface.launch(share=True)